ACIS door failure investigation: technical details and mitigation ...

186 

ACIS Door Failure Investigation: Technical Details and Mitigation Procedures 

Mark A. Kahan Neil L. Tice , William A. Podgorski Paul P. Plucinsky , & Keith B. Doyle * 

Abstract 

NASA's Chandra X-ray Observatory (formerly AXAF) was launched on July 23, 1999 and is currently in orbit 

performing scientific studies. Chandra is the third of NASA's Great Observatories to be launched, following the 

Hubble Space Telescope and the Compton Gamma Ray Observatory. One of four primary science 

instruments on Chandra, and one of only two focal plane instruments, is the Advanced CCD Imaging 

Spectrometer, or ACIS. The ACIS focal plane and Optical Blocking Filter (OBF) needed to be launched under 

vacuum, so a tightly sealed, functioning door and venting subsystem were implemented. The door was opened 

two and one-half weeks after launch (after most out-gassing of composite materials) and allowed X-rays to be 

imaged by the ACIS CCD's in the focal plane. A failure of this door to open on-orbit would have eliminated all 

ACIS capabilities, severely degrading mission science. During the final pre-flight thermal-vacuum test of the 

fully integrated Chandra Observatory at TRW, the ACIS door failed to open when commanded to do so. This 

paper provides a somewhat technically expanded description of the efforts, under considerable time pressure, 

by NASA, its contractors and outside review teams to investigate the failure and to develop modified hardware 

and procedures which would correct the problem. Although, despite considerable effort, the root cause of the 

test failure was never explicitly identified, the expanded technical information and factors presented here may 

prove of use to others working the details of door design. On ACIS our efforts ultimately became focussed on 

hardware and procedures designed to mitigate the effects of the potential, but unproven, failure modes. We 

describe a frequent real-world engineering situation in which one must proceed on the best basis possible in 

the absence of a complete set of facts. 

Introduction 

The Advanced CCD Imaging Spectrometer (ACIS) is one of the four primary science instruments on the 

Chandra (formerly AXAF) X-ray Observatory and one of only two focal plane instruments. A picture of the 

ACIS instrument is shown in Figure 1 . In this view, the X-rays would be imaged by the X-ray mirrors, located 

far above the instrument, and directed downwards into the detector housing, where the CCD imaging chips are 

located. The ACIS door is shown in the open position in the picture. Due to the presence of a fragile optical 

blocking filter over the CCD's and the need for extreme cleanliness, the ACIS focal plane must be kept in an 

evacuated detector housing and launched under vacuum. The blocking filter essentially "blocks" any visible or 

ultraviolet light from entering ACIS's aperture. Acoustic loading during the launch environment due to presence 

of air in the detector housing would destroy the delicate filters. The ACIS door seals the detector housing 

during ground operations leading up to launch. STARSYSTM paraffin (wax) actuators are used as components 

in a rotary actuator, to open and close the door for ground testing and on-orbit operations. 

The ACIS door mechanism, designed and built by Lockheed-Martin Co., was fully qualified and life tested (on 

an engineering model) and the flight unit had been actuated at least 23 times during component testing and 

during system assembly and test operations. However, during the final Chandra Observatory level 

thermal/vacuum test at TRW, the door failed to open when commanded to do so. This failure occurred in June 

of 1998, at a time when the Chandra launch was scheduled for early CY 1999. A similar failure of the door to 

open on-orbit would eliminate the use of the ACIS instrument and thereby greatly degrade mission science 

capabilities. 

Needless to say, this failure at such a critical time in the Chandra Integration and Test cycle was extremely 

distressing and generated an immediate, overwhelming and ultimately successful response from a combined 

team of NASA-MSFC, contractor and independent review team person ne112. 

* The work at Optical Research Associates was performed under ADF contract 97-6135. 

** The work at Lockheed-Martin was performed under MIT contract SC-A-I 24624. 

*** This work was supported at SAO by NASA Chandra contract NAS8-39073. 

Optomechanical Engineering 2000, Mark A. Kahan, Editor, 

Proceedings of SPIE Vol. 4198 (2001) © 2001 SPIE · 0277-786X/01/$15.00 

Downloaded from SPIE Digital Library on 09 Sep 2010 to 128.103.149.52. Terms of Use: http://spiedl.org/terms

Figure 1 ACIS Instrument 

Downloaded from SPIE Digital Library on 09 Sep 2010 to 128.103.149.52. Terms of Use: http://spiedl.org/terms 

Proc. SPIE Vol. 4198 187

188 

Design Requirements 

ACIS Door Mechanism Design 

The design of the ACIS Detector Assembly was performed by Lockheed-Martin Astronautics under contract to 

MIT (contract # SC-A-124624), who in turn was under contract to NASA-MSFC for the ACIS Instrument. 

Design work started in 1994 with a prototype unit built prior to Preliminary Design Review (PDR). A 

qualification unit was built and tested in early 1995 with the flight build and test starting in early 1996. An 

Engineering Unit was also built and was delivered to MIT in early 1995. There were many requirements 

which contributed to the final design of the ACIS Detector Assembly Door Mechanism. The key ACIS door 

requirements are listed below: 

I . Design Life 253 Cycles 

2. Operating Temperature Range 

-60°C to +45°C (Qualification) 

3. Flight Operating Temperature 

4. Survival Temperature Range 

20°C 

-76°C to +45°C 

5. Operating/non-operating Pressure 800 TORR to 0 TORR 

6. Operating Voltage Range 

24 Volts DC to 36 Volts DC 

7. Actuator Mm. Operating Torque > 5.1 Nm (45 ln-lbs) 

8. Actuator Mm. Stall Torque > 6.8 Nm (60 In-Ibs) 

9. On-orbit Operation 

10. Ground Operation 

One open cycle 

< 100 Cycles 

I I . Cleanliness Requirements Class IOOA and meet the Requirements of 

MSFC-SPEC-1238 and MIL-SPEC-1443 

12. Internal Vacuum Up to one atm Pressure difference Across Door 

13. Seal Capability 

Contamination control was a significant design driver for the door mechanism assembly and so parting 

compounds were not utilized and materials and components were selected which were known to be 

compatible with these requirements. The requirement for an internal vacuum (in the detector housing) for all 

ground processing and launch environments, combined with the very low allowable leak rate for such a small 

instrument volume, demanded a nearly perfect hermetic door seal. An unlubricated Viton ® 0-ring was 

chosen as the sealing material because of its vacuum compatibility and low outgassing properties. The torque 

requirements for the door actuator were derived from some early development tests which showed a 

requirement for at least 2.3 Nm (20 in-Ibs) of torque capability from the actuator when using a dry Viton 0-ring. 

Actuator trade studies performed early in the PDR design phase resulted in a decision to proceed with a 

Rotary Paraffin Actuator built by STARSYS Research to drive the door mechanism open and closed. This 

actuator (Part # HL4570B), based on a design from a previous NASA program, had redundant built-in limit 

switches and heaters with a positive latch at both extremes of travel. The minimum stall torque capability for 

this actuator was 6.8 Nm (60 in-Ibs), which was a factor of 3 greater than required. 

Door Mechanism Actuator and Linkaqe 

The rotary actuator uses two linear paraffin actuators, one for each rotational direction. Upon actuation, 

heaters inside a cartridge warm the internal paraffin. As the paraffin changes phase at around 70°C its volume 

increase forces a steel pin outward; this linear travel is converted into rotary motion of the output shaft and 

coupled ACIS torsion driveshaft. The driveshaft and a folding two-bar linkage rotate the door approximately 

92 degrees and out of the field of view upon opening; in the other direction the linkage is driven to a nearly 

straightened locking position when the door is closed. The linkage contains an adjustable turnbuckle, set to 

provide a calibrated force against the Viton 0-ring. This force, determined through development testing, 

provides adequate sealing for evacuation of the instrument without additional applied forces. A "swing link" 

supporting the door hinge axis allows the door to seat fully on the 0-ring so that 0-ring compression is uniform 

upon preload and evacuation. An internal shear disk in each actuator prevents excessive pressure inside the 

actuator from causing a catastrophic failure of the paraffin actuator cartridge & protects the other components 

in the device from damage. However, a shear disk rupture does render the actuator permanently disabled. 

Figure 1 shows the ACIS door mechanism in the open position, with the access cover removed and the 

Proc. SPIE Vol. 4198 


Door Mechanism Actuator and Linkage, Concluded 

STARSYS actuator on the right side of the photograph. The linear actuators are the cartridges extending from 

the top and bottom of the rotary actuator case. 

Actuator Torque Requirement 

The output torque required from the door actuator was almost entirely driven by the expected stiction that 

could develop in the 0-ring door seal. A cross-section of the mechanism is shown in Figure 2 and shows the 

open and closed door configurations. The two bar linkage applies around 222 N (50 lbf) of force (1.8 Nm or 

16 in-lbs) output torque from the rotary actuator) to the 0-ring seal when in the straightened position. This 

force is required to compress the 0-ring sufficiently so that the detector assembly can be evacuated. The 

mechanism itself consists of the aluminum door, dry Torlon ® bushings in all rotating joints, titanium/aluminum 

linkages, a titanium collimator, aluminum seal retainer and the 15-5ph stainless steel linkage driveshaft. The 

slip-fit Torlon bushings all have redundant moving interfaces which would nominally still allow rotation if one of 

Collimator 

0-ring Seal 

Seal Retainer 

Figure 2 - ACIS Door Mechanism 



190 

Actuator Torque Requirement, Concluded 

the surfaces were to bind. Running torque for the mechanism was very low, even in IG environments, since 

the door weighs only about 0.1 1 kgm (025 Ibs). The total running torque for the drive train was only around 

0.5 Nm (4.5 in-Ibs), which was mostly the frictional and spring losses inside the rotary actuator. The frictional 

losses in the Torlon bushings were minimal. Analysis and testing showed that there was not an issue with 

clearance changes that could result in any binding of mechanism components, even at the most extreme 

temperatures. Additionally (post-failure), the mechanical advantages at all joints were re-computed allowing for 

worst case angular positions (as loose tolerances can sometimes lead to unfavorable piece-part mechanical 

positions). This analysis showed no positions resulting in abnormal forces or restraints. 

Design Redundancy 

As is the case with most mechanisms or structures, the door mechanism drive train components connecting 

the rotary actuator to the door (driveshaft, linkages, etc.) were not redundant. Rotation joints in all linkages 

were intended to be single fault tolerant due to the inclusion of floating Torlon bushings. Adding a second 

rotary actuator would have significantly complicated the mechanism. The paraffin actuators themselves 

contained dual sealing 0-rings and redundant heaters. The rotary actuator also contained redundant limit 

switches. Return springs inside the rotary actuator were designed to low operating stress levels to avoid 

redundancy concerns. During the design phase, a single paraffin actuator with redundant heaters was 

deemed to be adequate for a one-time on-orbit actuation requirement. These actuators had significant flight 

heritage with extremely high reliability, and required only 28 volt input power to one of the redundant heaters to 

perform as designed. All of the electronics required to power the actuators were fully redundant and met the 

single fault tolerance requirement. 

Paraffin actuator shear disk rupture from high internal pressure was considered a potential single point failure 

resulting in loss of mission. During operation, the actuator reaches a temperature considerably above the 

paraffin melting point. If the case temperature becomes too high before heater shutoff, internal pressure is 

relieved by shear disk rupture. STARSYS research data showed a disk rupture threshold temperature of 

154°C; the design team concluded that a shut-off of 135°C left acceptable margin for door operation, based on 

our development testing. To protect against failure, redundant temperature sensors were installed onto the 

actuator case and were monitored autonomously by the ACIS Power Supply and Mechanism Controller 

(PSMC). Power was removed if the actuator ever reached 133-135°C. The subject of shear disk rupture will 

be discussed in more detail later in the paper. 

Qualification Testinq 

Pre-Flight Testing 

Pre-flight testing was performed per NASA procedures to fully qualify the mechanism for flight. These tests 

included a 253 cycle thermal/vacuum mechanism life test on a special life-test unit (CAMSIM), 0-ring stiction 

tests, acoustic tests, and random vibration tests. Initially it was felt that there were no significant issues that 

were not addressed in the final flight design for the door mechanism. One of the minor problems addressed 

was the burn-out of one of the life-cycle test heater elements. The root cause of this failure was a process 

change in a Minco heater that led to a delamination of the heater and subsequent burn-out as paraffin worked 

its way into the heater and fatigued the element. For flight, new heaters were fabricated with the original 

process and were re-qualified and life-cycle tested. A second problem occurred during the life-cycle test due 

to I C affects and rapid cycling of the mechanism and paraffin actuators. Convection inside the hot wax (only 

present in I G) resulted in a non-uniform re-freezing of the paraffin as it passed through the phase-change 

temperature. Since the paraffin froze at the bottom of the actuator first, the actuator output pin was prevented 

from fully retracting. This problem was addressed by procedural changes, a new retraction spring which had a 

higher spring constant, and an additional limit switch capability which would alert operators if the shaft was not 

fully retracted after a close cycle. There was a procedure in place, which never had to be used, to re-set the 

actuator should this problem repeat. These qualification tests showed that the mechanism was ready for flight 

build with no major design changes. 



Flight Unit Testing 

The flight unit Detector Assembly door was opened and closed at Lockheed-Martin at least 12 times as part of 

the component level Protoflight Test program under various test conditions prior to delivery to MIT. Testing 

included pre and post environmental exposure performance tests, random vibration tests, and thermal vacuum 

tests. The flight unit Detector Assembly was delivered to MIT in early 1997 and the flight focal plane and 

optical blocking filters were installed in April of 1997. The door was cycled open and close I I more times in 

system level testing as described below. 

An ACIS system level Thermal Vacuum test of all of the flight hardware was conducted at MIT Lincoln Labs. 

The door was cycled two times as part of the long and short form functional testing of all ACIS hardware, at 

hot and cold temperature extremes, with nominal results. At the completion of this test, the integrated ACIS 

was delivered to MSFC for an additional one month calibration test in the MSFC X-Ray Calibration Facility 

(XRCF) thermal/vacuum chamber. During X-ray calibration, several additional door operations cycles were 

performed under flight-like operating conditions with nominal performance. After calibration activities were 

complete, the unit was vibration tested as part of the final acceptance test by MSFC. Pre and post test 

mechanism performance, which included door cycling, was nominal, thus adding four more door open/close 

cycles. The integrated ACIS was then shipped to Ball Aerospace in Boulder for integration onto the Integrated 

Science Instrument Module (ISIM). The Detector Assembly was kept under vacuum for 5 months until the 

ISIM was fully integrated. ISIM level thermal/vacuum testing started in October 1997. During these tests the 

door was opened and closed two more times under flight-like conditions. One additional successful door 

opening occurred at ambient temperature and pressure in January of 1998. The ISIM was then shipped to 

TRW for integration onto the Chandra telescope. Periodic pump downs of the ACIS Detector kept the internal 

pressure between 0 and 10 torr for all system integration and test activities (including system Acoustic test). 

Related Testiq 

Another STARSYS HL4570B rotary actuator (same model as for door) was used in the ACIS Large Vent Valve 

mechanism. A failure of this actuator occurred during the ACIS system thermal vacuum test at MIT Lincoln 

Labs in April of 1997. This failure was due to an operator error which caused both the open and close 

actuators to be powered simultaneously. Under this condition, one of the shear disks ruptured and required 

the removal and replacement of the Large Vent Valve rotary actuator with a flight spare actuator. Corrective 

action was to follow procedures to prevent simultaneous powering of open and close actuators. The large 

vent valve was cycled hundreds of times during test and integration and was never an issue before or after this 

event. However, this failure on the large vent valve pointed to a potential weakness in the actuator design. As 

discussed earlier, disk rupture will normally occur at a case temperature of about 154°C, and the PSMC will 

cut-off power when the actuator temperature reaches about 135°C. However, it was discovered that this 

154°C rupture temperature was based on a linear actuator in which the output pin is allowed to extend 

normally to full extension. Therefore, the rotary actuator must operate with stall torques less than 6.8 Nm 

(60 In-lbs) on the rotary output shaft. if the output pin is restrained from extending, then there is no additional 

volume for the paraffin to expand into and the shear disk will rupture at a much lower temperature if power is 

not removed promptly. Thus, in the case of a restrained actuator, the temperature sensors do not provide 

protection from rupture except at end-of-stroke. 

As it turns out, instead of a constant temperature shut-off point, each paraffin actuator could be protected by a 

variable shut-off temperature dependent on the amount of linear extension of its output shaft (transformed into 

rotary actuator shaft rotation). At zero degrees of rotation for the rotary actuator mechanism, the shut-off 

temperature should have been around 90°C rising to 150°C at full rotation or end-of-stroke. However, since a 

normal operation of the door or large vent valve would result in an actuator temperature of I 05-1 1 0°C, we 

could not simply lower the shut-off temperature without affecting the ability to operate normally. It was also too 

late in the program to re-design the electronics to provide a variable shut-off temperature. It was evident, 

however, that in the case of a bound actuator shaft early in the stroke, we would not achieve our goal of 

opening the door. This potential new protection would have allowed us to try opening the door many times 



192 

Related Testing, Concluded 

rather than just once. The decision not to implement this protection was based on our high confidence n our 

ability to open the door, since we had never observed a stiction torque above 2.3 Nm (20 in-lbs) in over 

3 years of testing. 

To summarize the component and system level testing: the ACIS flight unit door was successfully cycled at 

least 23 times prior to the opening attempt at TRW under various operating temperatures and pressures. The 

ACIS team had confidence in the door mechanism performance for system level thermal vacuum test and for 

flight. There had been no anomaly in any of the previous door opening attempts. In fact, closing the door had 

always been of greater concern since the door was closed against a frozen (stiff) 0-ring several times during 

these tests. Measured paraffin actuator case temperatures were always higher when closing against a frozen 

0-ring, but the unit was qualified to perform under these conditions so it was never an issue. 

Door Opening Failure 

Door Opening Failure in System Thermal/Vacuum Test 

The fully integrated Chandra X-ray Observatory was subjected to over one month of thermal/vacuum testing at 

TRW (Chandra prime contractor). The final test planned for this period was a stray light test, in which the 

ACIS door would be opened and the ACIS used as a sensor to measure stray light levels within the telescope 

(lamp banks simulating the sun were installed in the TN chamber). Test conditions prior to the stray light test 

were nominal for the telescope, with the ACIS running at cold temperatures as would be normal for 

observations in flight. Since there had been several past door openings with ACIS cold and a warm-up would 

have taken extra time, it was decided to open the door in the cold condition. At the time, there was no 

objection from anyone on this point. 

The ACIS door opening procedure was run at 5am PDT on June 18, 1998, but the door did not open. The 

information available at the time was as follows: 1) Door Open Actuator exceeded its power shut-off 

temperature at 133°C with autonomous removal of power by the ACIS electronics (PSMC), 2) using X-ray 

measurements from a door mounted calibration source, it was determined that little to no motion of the door 

had occurred, 3) STARSYS actuator limit switch operation indicated less than 20 degrees of rotation had 

occurred, 4) ice had been observed on external surfaces of the Observatory, indicating the presence of at 

least some water in the TN chamber, 5) the ACIS door and camera body were at —-60°C, and 6) the 

STARSYS Actuator was at -42°C at the start of the procedure. 

Initially Suspected Failure Modes 

Post-test Failure Recovery and Analysis 

As is usual in such cases, a small "tiger" team was formed to determine an immediate course of action. Even 

at this early stage, the team believed there was a high probability that the ACIS would have to be removed 

from the Observatory. A full stall of the mechanism with no door movement would more than likely have 

ruptured the shear disk inside the Paraffin Actuator based on experience with the Large Vent Valve 

Mechanism. A fault tree was created to help determine possible causes of the anomaly. The final form of the 

fault tree is shown in Figure 3. The first three (non italicized) items were those which were initially considered. 



Initially Suspected Failure Modes, Concluded 

A. Door/Seal Retainer Bound 

. Cold weld 

. Ice 

. SealAdhesion 

- Electrostatic 

- Djffusion 

- van der Waals 

- ChemicalBonding 

B. Rotary ActuatorBound 

I Parts Bound 

. Mismatched parts 

. Contamination 

C. Paraffin Actuator Bound or Failed 

. Heater failed 

S Wax leak 

. Start low temp 

. Low lube 

D. Rotary Actuator Shear Disk Previously Ruptured 

E. Test Procedure not followed 

F. Retarding Torque > 6.8Nm (60 in-lbs.) 

. Over-center 

. Seal Separation 

G. Linkages Bound 

. CTE Mismatch 

. Bushing/Linkage bending 

. Ice 

. Contamination 

Figure 3 - ACIS Door Failure Fault Tree 

Ideally, the team would have liked to see the unit left undisturbed to prevent the loss of information that could 

ultimately lead to determination of the failure mode. However, the MIT and Lockheed-Martin team concluded 

that there was too great a risk of damage to the focal plane and OBFs from contamination if the door were not 

fully closed prior to re-pressurization of the chamber. The door provides a nearly hermetic seal which would 

protect the focal plane and OBF's from contamination as the shrouds and spacecraft were warmed up and the 

chamber re-pressurized. 

Recovery from Failure in ThermalNacuum Test 

The decision was made to warm the detector assembly and door to 25°C using heaters that were designed for 

baking off contamination on-orbit. To protect the instrument and get the door closed, the open actuator had to 

be heated to soften the paraffin. With a warm open actuator, there was a higher probability that the close 

actuator could re-set the mechanism and latch the door closed again. The procedure was performed at 7pm 

PDT without success. Neither the open nor the close limit switches changed state, and the close actuator 

went over-temperature at 133°C as the PSMC autonomously removed power. A nominal opening or closing 

would normally result in an actuator temperature of around 105°C at limit switch shutoff. Now the question 

was whether or not the door was sufficiently sealed to allow for chamber re-pressurization. As dry nitrogen 

was bled into the chamber, the differential pressure transducers on the detector assembly quickly confirmed 

that the door was sealed, since a vacuum was measured within the detector housing. The focal plane and 

OBF were in a safe condition, since one atmosphere of pressure differential provides about 500 pounds of 



194 

Recovery from Failure in ThermalNacuum Test, Concluded 

closing force on the door. Therefore once the door sealed, a positive latch was not necessary and the 

instrument could be handled safely without the fear of damage to the focal plane and OBFs. 

During the next several days Chandra was removed from the chamber, the ISIM was removed from the optical 

bench and partially disassembled, then the ACIS was removed for shipment back to Lockheed-Martin. Initial 

external inspections at TRW looked normal except that the door did appear to have metal to metal contact with 

the seal retainer. This metal-to-metal contact could support the cold welding hypothesis but later inspections 

and testing would discount this failure mode. The ACIS detector was then shipped back to Lockheed-Martin 

where inspections and disassembly could occur in a class 100 clean room. 

Initial Inspections at Lockheed-Martin 

Failure investigation efforts centered on closing branches of the fault tree shown in Figure 3. Initial inspections 

of the door mechanism at Lockheed-Martin did not reveal the cause of the failure. All linkages appeared to be 

free and the two bar link was slightly over-center, which explained why the door was able to seal. It appeared 

that the door was metal to metal with the seal retainer, but until the detector was re-pressurized, we would not 

know if the door and seal were bound to one another. Since there was still a vacuum inside the detector, as 

measured with the Vacuum Ground Support Equipment (VGSE), we elected to remove the STARSYS actuator 

prior to detector housing re-pressurization, so that it could be inspected independently of the rest of the 

mechanism. 

Door Seal/Retainer Bound (For the Non-Italicized Fault Tree Item A) 

Since cold welding of the gold plated door to the aluminum seal retainer was a possible failure mode, extreme 

care was taken as the detector assembly was re-pressurized. Dial indicators were mounted above the door to 

measure movement of the door as dry nitrogen was allowed back into the camera body. The door slowly rose 

off of the seal retainer as the internal pressure was equalized to the ambient pressure in the clean room. The 

0-ring was still compliant and relaxed as the external pressure was removed from the door. Cold welding 

and/or galling was ruled out both because the 0-ring and its gland had been sized to preclude line-to-line 

metal contact after a compressive set, and because the as-built 0-ring had enough spring force to push the 

door away from the seal retainer. Close inspection revealed no signs of cold welding, metal burnishing, or 

metal-to-metal oxide/contaminant adhesion. This also would have been the case when the TRW thermal 

vacuum chamber was initially evacuated since the pressure differential across the door would been reduced in 

a similar manner. 

With the differential pressure equalized, the door could be opened and running torque and seal adhesion could 

be measured. Using a torque wrench to drive the door open, the seal adhesion plus running torque was 

measured to be 2.6 Nm (23 in-Ibs) which was near the expected value. Seal adhesions had never been 

measured above 2.3 Nm (20 in-Ibs) and with a running torque of around 0.5 Nm (4 in-Ibs), the seal adhesion 

was 2.2 Nm (1 9 in-Ibs). The full capability of the STARSYS actuator as measured in subsequent testing, was 

around 10.2 Nm (90 in-Ibs) which gave a factor of safety of over 4 above what was measured. With running 

torque within normal specifications, no problems were found with the mechanism. Further disassembly and 

inspections showed all dimensions for bushings and other components to be within specifications at room 

temperature. No binding or interference was found which could have contributed to the failure. Additional 

testing on the Torlon® bushings showed that the coefficient of thermal expansion was as expected and that 

analytically there were no concerns with the mechanism binding at low temperatures. 

The over-center torque on the door mechanism was also found to be within specifications at about I .9 Nm 

(17 in-Ibs). Since this flight mechanism was set up in an over-center condition, due to normal rotational and 

machining tolerances, there was concern that if the over-center torque was too high then it could have 

combined with another failure mode to result in excessive retarding torque. This was shown not to be the case 

with testing on other engineering units even for an incompressible 0-ring. Conservatively, the flight unit was 

re-assembled with the mechanism adjusted in a slightly under-center condition by building a new adapter 

bracket. 



STARSYS Actuator Inspection and Test( Fault Tree Items B and C) 

Failure of the STARSYS rotary actuator was one of the failure modes identified on the fault tree. X-ray 

inspection of the unit did not show any anomaly. As was expected, the open actuator shear disk had ruptured 

due to excessive pressures inside the actuator. The open actuator output shaft was slightly extended which 

explained why the latch and limit switches were not engaged. Some drive shaft rotation was also observed in 

the X-rays of the actuator. There was no evidence of any foreign material or debris that could have jammed 

the mechanism. The open actuator cartridge was then removed from the actuator assembly. As the actuator 

was removed, the latches and limit switches fell into place. The slightly extended output shaft, which was 

frozen in the paraffin, had prevented the actuator from re-setting completely. The shear disk looked like a 

normally ruptured disk with no signs of any material weakness. Material properties were measured for the 

shear disk and were as expected. There were no signs of wax leakage and internal inspections of the wax 

cartridge were nominal. Force measurements were made to determine the running torque for the rotary 

mechanism and were nominal at both room temperature and at —45°C. 

Prior to total disassembly, the close actuator cartridge was removed and installed onto the open actuator side 

of the rotary mechanism assembly. At STARSYS, the unit was tested to failure with a full stall to characterize 

the performance of the actuator and shear disk under these conditions. This full stall test, performed at -45°C, 

simulated the condition where the door mechanism did not rotate and all bushings and linkages were frozen 

solid. The temperature profiles were similar to those which were measured at the time of the failure up until 

the actuator reached a temperature of around 129°C, at which time the shear disk ruptured. This is slightly 

lower than the estimated failure temperature of 132°C at TRW, but would be expected with this more severe 

test condition. The door mechanism linkages and torsion shaft would have allowed more rotation of the rotary 

actuator and thus a higher temperature at shear disk rupture. It is interesting to note that both shear disks 

were nearly protected since the PSMC shut-off temperature was set at I 33-1 35°C. The peak torque at shear 

disk function was 10.4 Nm (92 in-Ibs) vs. the 6.8 Nm (60 in-lbf) specification. Material testing showed that both 

the open and close shear disks were similar in thickness and shear properties, so it is likely that the STARSYS 

rotary actuator output was around 10.2 Nm (90 in-Ibs) at the time of the failure at TRW. Complete 

disassembly and inspection of the mechanism and actuators did not reveal a problem. Nothing was 

obstructing the mechanism and the only discrepancy was the ruptured shear disk. This closed out numerous 

branches on the fault tree since everything appeared and functioned nominally. 

The shear disk rupture point varies linearly depending on the amount of extension of the actuator linear output 

shaft, which can be anywhere from 0 to 19 mm. At 0 mm of linear extension, test data showed the actuator 

would reach an equivalent pressure to achieve 6.8 Nm (60 in-lbf) on the output shaft at around 90°C. The 

rupture temperature would be higher at an equivalent output torque of I 0.2 Nm (90 in-Ibs). For an extension of 

19 mm, the shear disk would rupture at an estimated 154°C. Subsequent discussions will show that we 

believe the flight unit shear disk ruptured at around I 32°C which also supports the evidence that the door 

mechanism had rotated 20-25 degrees (5.7 mm of linear extension) at the time of failure and was not locked 

up. The clearances in the bushings and the windup in the mechanism components could allow the actuator 

linear output pin to extend as much as 5.7 mm even with a stuck door. 

Failure Investigation Team 

NASMndustry Team Formation and Failure Investigation Approach 

The small team of Lockheed-Martin, NASA-MSFC and STARSYS personnel as discussed above initiated door 

mechanism failure diagnostics. Their goal was to determine the cause of the failure, fix the problem, and then 

return the ACIS detector to flight configuration so that it could be reinstalled on Chandra and still meet the 

(then projected) January 1999 launch schedule. This team was initially given two weeks to complete these 

activities, based on the (optimistic) assumption that the failure mechanism would be easily identified and 

rectified. As it eventually became clear that the failure was not going to be quickly resolved, NASA-MSFC 

formed a larger ACIS Door "Tiger" team which included members from NASA-MSFC, selected members the 

Chandra External Independent Readiness Review (EIRR) Board (from Optical Research Associates & the 

Aerospace Corporation), MIT, Lockheed-Martin, STARSYS, TRW and SAO. The team's function was to 

oversee and facilitate the ongoing ACIS door failure investigation and the decision making process. In the end, 

the team was to operate for almost one year, as circumstances dictated, with weekly telecons and meetings as 

necessary. 



196 

Failure Investigation Team, Concluded 

The situation at the time the larger team was formed, in early July 1998, was not a promising one. As 

described above, initial work by the smaller team had ruled out many of the proposed failure mechanisms. 

Also, program schedule pressure was large, even though the launch date had been slipped to mid-summer 

1999 due to other factors. The expanded team, along with senior NASA and program management, decided to 

pursue parallel paths of off-line failure investigation along with the re-build and re-test of the ACIS detector and 

then re-integration with the Chandra Observatory. It was believed that the re-test effort would also supply 

valuable data to the failure investigation team. As a further backup possibility, NASA-MSFC even developed 

an alternate actuator that could have been substituted into ACIS should subsequent problems with the above 

paths occur downstream closer to launch. 

With more resources available now, many activities could proceed in parallel, even though the primary activity 

remained at Lockheed-Martin. These activities included the expansion of the fault tree to include a more 

detailed examination of the italicized items some of which were not initially considered, further investigation 

into the issue of ice formation, more detailed investigations into the stiction properties of Viton, a search of the 

prior test data to look for a possible inadvertent powering of the door actuator, development of a detailed finite 

element model of the door mechanism and 0-ring to provide additional design verification and, of course, the 

re-build of the flight ACIS detector for a re-test at TRW in August of I 998. 

Further Failure Investigations (Italicized Items, SectionA, Figure 3), Changes, and Launch Liens 

The expanded review team briefly reinvestigated a large number of fault tree branches that had been 

previously closed by analysis, test, and inspection. Possibilities reviewed included considerations that slow 

voltage ramp-ups in testing could have inadvertently produced a condition that might burst an actuator, as well 

as microscopic effects of ice/water, &/or Viton cold-flow. Additional tests and analyses were conducted in each 

of these areas, and selected design and operational changes were made to address each open fault tree 

branch. Mitigation schemes were implemented and verified by tests &/or FEM techniques including: (a) the 

use of a controlled energy slug and potentiometer to allow multiple opportunities to open the door on-orbit 

(reducing any paraffin actuator shear disk operational risk), (b) the previously noted elimination of "overcenter" 

torque so as provide additional margin, and (c) use of door grounding to eliminate concerns relative to 

any stiction induced by electrostatic forces. Although a few KV were initially computed (given an Ion 55 source 

which was attached to the door for later use in on-station telescope calibrations), this value was later reduced 

to low levels as both further analysis and testing showed that the small gap at the seal interface can't support 

a significant level of charge build up. Regardless, the grounding strap was added for safety. 

It was believed that a warm door opening on-orbit would close all open fault tree branches, but a launch 

constraint was imposed requiring a successful KSC open/close cycle. Further, an additional defacto launch 

constraint was imposed which required successful Engineering Unit testing using flight-like timelines for door 

closure. Here, the Engineering Unit (EU) testing incorporated both expected vibration & acoustic loading as 

well as use of selected on-station heating as part of a thermal 'mitigation sequence" (this sequence will be 

discussed shortly). It was decided that the launch would be scrubbed if EU door opening required over 45 in-lb 

of torque (laboratory testing showed I 9 in-lb to be a nominal stiction value, with worst-case test data of 44 in-lb 

achieved only once at -60°C after 28 hours). Recall that 90 in-lb was needed to cause the actuator to fail, so in 

constraining torques to 45 in-lb there would still be about 2X torque margin. While assessing a value to use as 

an EU toque limit, we also linearly added all the identified non-seal-surface-based stiction sources, and 

determined that only 55 in-lb of seal-based resistance might be sufficient to fail an actuator. 

The cracking of the door seal at KSC would reset the stiction clock, but this was not without some risk of its 

own. Though the contamination effects of a KSC opening were evaluated and found to be acceptable, there 

existed the possibility that a never-seen ambient-only open failure might occur given the potential of van der 

Waals forces. (Recall how difficult it is to pull apart two mating microscope slides, especially with an 

intermediate thin film of water. Further, try and remove an ice-cube from a countertop after the cube has 

begun to melt. This type of stiction ties to the electrical polarization induced by surface particles; forces go 



Further Failure Investigations (Italicized Items, Section A, Figure 3), Changes, and Launch Liens, Continued 

inversely as the seventh power of intermolecular distances involved.) Since ambient open forces were 

generally 2X those in vacuum, there seemed to be reason for suspicion. Also, since van der Waals forces can 

climb to 20-30 Ksi and the door-seal is approximately 20 inches in overall length by 0.1 inch in compressedcross-section, 

stiction forces had the potential to overcome the available torque, even if only 1/20% of the 

mating surfaces were involved. The Team recognized that should a water-film based stiction actually occur at 

KSC, the case that the door would open on-station, despite the closure of fault tree branches, would be further 

weakened, and the launch might well be jeopardized. Moreover, any such ambient failure would be hard to 

"mitigate" at KSC as adequately dehumidifying the area would be quite difficult. Also, only small thermal 

changes could be introduced at ambient conditions, and small changes were determined to be insufficient to 

provide effective stiction mitigation. The Team decided to accept this risk, and press-on. 

In parallel to firming-up KSC plans, the team developed procedural details & a flow-chart for a warm on-station 

opening that incorporated torque relationships as a function of time and temperature, and which accounted for 

all possible hardware tolerances (including issues of timing errors, temperature read-out anomalies, etc.). 

Also, a simulator (the CAMSIM) was used to check cold torque's, and this unit was configured to act as a 

Flight Unit Shadow. The CAMSIM would then be able to act as a realistic test-specimen should on-station 

difficulty eventually be encountered. 

Transfer function calculations and tolerancing of all linkages were reworked, and all build-books were 

reviewed. No toleranced positions yielded sufficiently high torque to produce failure, even given the potential 

mechanical interferences in worst-case error stack-ups. These stack-ups accounted for temperature changes 

under differential expansion (Torlon/H20 testing gave 36-37 ppm/°C matching vendor data of 30 ppm/°C), as 

well as relative humidity swelling and changes in the Torlon bushings (where creep effects were also 

measured). A detailed review of displacements Vs angles Vs joint mechanical advantages confirmed that 

iced/bound joints give insufficient restraint to cause a failure ( 2 in-lbs of change). This was further confirmed 

by additional CAMSIM iced-Up testing. This work, taken together with our analysis, helped to eliminate linkage 

binding as the root cause of the failure. 

Our new on-orbit operating procedures utilized the application of pulsed power to help prevent 

functioning/bursting of the shear disk. This allowed repeated attempts to be made to open the door. This new 

approach was verified on the Flight Unit (8 times), and actuator temperature/torque performance was fully 

characterized so we would know what to expect on-station. Also, telemetry rates were increased during door 

operation so as to allow for real-time interaction and possible intervention during flight door opening. Flight 

opening temperatures were set at ambient levels, but mitigation plans were included as a formal part of our 

door opening procedure. Here the plan was to rapidly thermally cycle the seal interface from 20-25°C to -60°C 

and back to 20-25°C, as ground testing showed that this type of cycling resulted in a 90% reduction in 0-ring 

stiction. 

The ACIS instrument operates with the detector housing at -60 °C. Bakeout heaters are supplied to 

temporarily raise the detector housing temperature to 25 °C in order to remove molecular contamination 

collected by the cold surfaces in the detector housing. Under normal conditions the radiators can cool the 

ACIS detector housing from 25 °C to -60 °C in about I I hours. In contrast the bakeout heaters can raise the 

detector housing temperature to 25 °C in about 4 hours. For the sake of mitigation these temperature slopes 

constitute a rapid temperature cycle. On orbit, the bakeout heaters were on when the ACIS door was opened, 

raising the door seal to 20-25 °C. Should the door have failed to open, the heaters would have been turned 

off and the temperature of the ACIS detector housing allowed to fall to about -60 °C. Immediately after 

reaching low temperature, the heaters would then have been turned back on and the detector housing warmed 

back up to about 25 °C. After reaching high temperature, the door would then have again be commanded to 

open, as experiment has shown that for adhesions formed at 25 °C, this technique reduces residual adhesions 

to insignificant levels. 



198 

Further Failure Investigations (Italicized Items, Section A, Figure 3), Changes, and Launch Liens, Continued 

As work on the opening sequence proceeded, testing and forensics continued. Black "lines" were seen on the 

parted sealing surfaces. Surface reflectivity shifts due to changes in the stociometry of the gold surface coating 

were initially suspected, but forced 0-ring extractions and lR Spectroscopy resulted in identification of the 

potential that a compressed baked Viton seal (two 3-4day 40+ °C cycles) could give rise to small amounts 

(

Further Failure Investigations (Italicized Items, Section A, Figure3), Changes, and Launch Liens, Concluded 

opposing sealing surface, the two surfaces would move laterally relative to each other by an amount greater 

than the average size of any surface pits that might be frozen in place. Allowing for Vitons cold elasticity and 

strength, it was speculated that, given this limited FEM analysis, the net effect of these thermally induced axial 

and lateral motions would result in a pulling apart and shearing of any bond formed at the interface by Viton or 

water ice. This proved true even with cold Viton peel forces & strength up by I .5X to 3X, based on limited 

testing. 

0-ring Finite Element Model 

assumed fixed, relative to gland, 

at sides and bottom; 

door effects excluded 

0-ring Motion Based on AT = -80 °C 

Lateral Motion X-Dir = 2 im w/r/t door 

Vertical Motion Z-Dir = 8 pm w/r/t door 

0-ring Motion Based on Temp Gradient of 

20°C between Doorand Gland 

Lateral Motion X-Dir = 35 im 

Combo of Effects May Break Stiction 

Pit Size: 10 tm wide; 2 im deep 

8 im Z-Disp due to Soak is > 2pm 

35 pm X-Disp due to Gradient is > 10 pm 

Figure 4 AXAF Door O..ring Motions 

Relative 0-Ring Z -Disp; AT = -80 °C 

Beyond these factors, our Team also looked in detail at the specific fabrication processes used to make 

Chandras 2-162 Viton 747-75 seals. In preparing Viton it is usually advisable to avoid alcohol wipes, though 

Viton s less damaged than many epoxies (e g Kapton has been rumored to experience apparent changes n 

surface texture) We also measured the material's diffusion and off-gassing rates (boiled 0-rings absorbed 

3 8% water by weight Vs 0 38% for standard unboiled 0-rings) Viton (Vinylidene Flouride & 

Hexafluoropropylene Copolymer in a formulation referred to as V747-75) contains carbon black high activity 

Magnesium Oxide (3%) & Calcium Hydroxide Hydrated MgO has the formula Mg(OH)2 and there are some 

Indications that MgO & Mg(OH)2 also stick to metals (especially the later) Chandra s seals were injection 

molded which results in a higher molecular orientation than compression molding It also adds mold 

contaminants and release agents into the mix. However, testing showed all these factors to be unlikely as root 

causes of the failure. 

We also surveyed other polymer bonding models Vs temperature in development at the Argonne National 

Laboratory, Georgia Tech., U. Cal. Santa Barbara, Pitt., Berkeley, Unitel Mixte de Recherches CNRS - Elf 

Atochem (France), and others, as well as tribiology studies on friction and wear. However, no additional 

potential root causes surfaced as a result of these reviews. 



200 

ACIS Detector Rebuild and Re-Test 

In the rebuild of the ACIS detector, much attention was given to possible ways to gain information on 

performance of the door actuator. Instrumentation to monitor door mechanism performance was added prior 

to the re-test which started in August, 1998. This instrumentation included non-flight strain gages to measure 

actuator output torque, a potentiometer on the torsion shaft (flight) and an additional temperature sensor to be 

used for "pulsed heating" of the paraffin actuator. This pulsed heating technique, in conjunction with 

potentiometer readings, would prevent the heating of the paraffin actuator so as to rupture the shear disk. 

Finally, a rapid warm-up and cool-down procedure was developed and demonstrated to reduce 0-ring stiction 

(one of the possible causes of the failure) in case the pulsed heating technique did not initially open the door. 

The re-test of the ACIS door mechanism (August/September 1 998) was intended to re-qualify the door 

mechanism under flight conditions. It was decided to re-test the ACIS door as mounted to the ISIM only, not 

the entire Observatory. This would greatly simplify the testing and also allow work on the Observatory to 

proceed in parallel with ACIS/ISIM re-test. A key decision of the failure investigation team and program 

management was that the test would take place with the detector housing warm, as was planned for the onorbit 

opening. This would eliminate any issues of ice during the re-test, but since water itself was an issue due 

to possible van der Waals forces, water was to be injected into the thermal/vacuum chamber to simulate the 

outgassing of the complete Observatory's composite materials. Some team members would have preferred to 

duplicate all test conditions at the time of failure (including detector temperature), but there was concern that 

another failure, even under non-flight conditions, would jeopardize program schedule. 

The re-built ACIS detector was shipped back from Lockheed-Martin to TRW in early August 1998, and was 

then integrated onto the ISIM. The lSlM with both Chandra focal plane science instruments (ACIS and HRC) 

was then moved into the TRW thermal vacuum chamber. The subsequent test then simulated as closely as 

possible the projected on-orbit door opening scenario. The predicted temperature profile was closely followed. 

The ACIS detector housing was kept cold until about 14 days into the test, at which time the door was opened. 

Prior to door opening, the detector housing temperature was raised to 25°C using the ACIS bake-out heaters. 

The door was opened (and then closed) a total of five times during the lSlM re-test, all using the pulsed 

heating technique. The first three openings. were done with a bus voltage of 28 volts and the last two at 

32 volts. The door angle potentiometer and driveshaft strain gages were monitored during these openings. All 

of the data from these openings was nominal. This successful re-qualification of the ACIS door mechanism 

allowed the program to proceed towards launch, but the fact that the cause of the earlier failure was still not 

exactly known was troubling to everyone. The ACIS failure investigation team would work for many more 

months, still attempting to discover the root cause of the earlier failure. 

It should also be noted that there were additional door openings in an ambient condition scheduled and 

executed to gain confidence in the door opening mechanism prior to launch. This was accomplished through 

the use of the VGSE to re-pressurize the ACIS detector housing cavity with nitrogen to equalize the pressure 

prior to opening. This was not originally planned due to the risk of contamination entering the cavity and of 

inadvertently damaging the thin spectral filters. However, a more detailed look showed that if procedures were 

carefully followed, these risks were small enough to be acceptable, given the circumstances. 

Ongoing Failure Investigations 

The expanded ACIS door fault tree is shown in Figure 3. This tree evolved over the course of the investigation 

and guided our efforts to identify the failure mechanism. As mentioned above, early post-failure inspections 

and tests addressed some of the branches of the tree. 

Pre-ruptured Shear Disk (Fault tree Item D) 

One possible failure mechanism which had not yet been ruled out, was an accidental activation of the open 

actuator under ambient pressure conditions prior to the TRW test. This would have taken place after the last 

opening at Ball (January 98) and before the opening attempt during Observatory ThermalNacuum testing. 

With a vacuum inside the detector, the shear disk would have ruptured if the open actuator were powered 



Pre-ruptured Shear Disk (Fault tree Item D), Continued 

when the Observatory was in an ambient environment, since the mechanism does not have enough torque 

margin to overcome one atmosphere of pressure against the door. Chandra flight telemetry was monitored 

anytime that Chandra or the ISIM were powered up. An investigation of the available telemetry data showed 

no time at which the actuator temperature had exceeded ambient. However, a potential failure mode was 

discovered when it was determined that when the spacecraft voltage was ramped up slowly, the ACIS 

mechanism controller could power up without receiving a command to do so. The team was able to duplicate 

this condition on the ACIS test hardware and was able to show that the mechanism could have been powered 

up inadvertently. However, the temperature profile and limit switch actuation measured at TRW during the 

door open failure did not match the characteristics which would be obtained for a pre-ruptured shear disk. 

(When the actuator is stroked, the expansion of the paraffin gives an additional tell-tale; it produces a 2 °C to 

3 °C thermal inflection point in valve time-temperature traces.) This conclusion is based on extensive testing 

with non-flight hardware and also the actual flight actuator used in the failed opening attempt. 

As part of this work the timing accuracy's of all commands were also reexamined (data logging within major & 

minor data frames), and this too was cleared as a potential failure source. Commands go to the controller at 

the start of major frames (within 3 - 4 seconds), while both door actuator and valve temperatures are recorded 

later, in minor frames. This type of signal timing initially made the door actuator appear to rise in temperature 

before data-tell-tales showed that the actuation command was actually sent. 

Figure 5 shows a typical torque curve. The strain gauge amplifiers were zeroed prior to opening even though 

this measurement contains significant zero offset. This offset is caused by the strain in the shaft required to 

maintain the 50 pound preload on the closed door. When the actuator unlatches (at -85 seconds in Fig 5) the 

forces from the preload back-drive the actuator until the running torque of the actuator arrests this motion. 

This exhibits itself as a level portion of the curve where the linear wax actuator "catches up" to the back-driven 

output shaft. During this time the limit switch picks up showing that the actuator is unlatched. As the wax 

actuator continues its stroke what remains of the 50 pound preload is removed as shown in Figure 5 by 

steadily declining torque. Eventually, the preload torque is completely relieved, which removes the 

compression on the door linkage. At this time the linkage travels through a dead band caused by the 

clearances in the various mechanical components. This too is exhibited by a second, level segment of the 

curve ( -1 00 seconds) and represents the true zero torque value at the shaft. As rotation continues, the 

linkage begins to pull on the door and the shaft torque curve in Figure 5 continues to decline. At this point the 

linkage is in transition from compression to tension and so represents a change of sign for the shaft torque. 

From this point on the torque measured is the torque caused by the adhesive forces between the 0-ring and 

the door and any small running torque caused by bearing friction and gravitational loads. As can plainly be 

seen this continues until the torque curve discontinuously jumps. This discontinuity has been visually 

correlated with door first motion on all of the tests performed. A small residual strain remains after the 

discontinuity and is a result of the now freely hanging door with gravity pulling it open and any residual torque 

in the mechanism. 

For the purposes of this analysis the second plateau caused by the unloading of the door mechanism is used 

as zero. The minimum reached just prior to the discontinuity was used as the opening torque. Thus, in the 

example shown in Figure 5 the opening torque is about 6.5 in-lbs. 

Microswitch triggering tests were also conducted (positions; effects of energy stored in the linkage, etc.). 

These tests showed that energy stored in the linkage can move switch positions by up to 0.05", effecting the 

apparent timing of critical mechanical events (the shaft rotates from I OO to 200 under 60 in-lb of torque). 



202 

Pre-ruptured Shear Disk (Fault tree Item D), Concluded 

U) 

.0 

C 

a) 

0 

2 

0 

-2 

-4 

-6 

-8 

-10 

-12 

-14 

-I 6 

-18 

80 

)-••••+••• ''+•- ••• 

3 

1 

CAMSIM Opening Torque 

45.5 Hour Vacuum Compression 

Ice on Mechanism (Items A and G of Fault Tree) 

—4-— Drive Shaft Torque (in-Ibs) 

— Switch I 

witch 3 

Figure 5. Typical Torque Curve for Opening ACIS Door 

Icing of the mechanism (as opposed to the door seal itself) was thought to be a promising failure mode since 

the failure occurred when the door was opened cold and frost was seen on the Observatory optical bench prior 

to the opening failure. However, analysis showed that ice could not have caused the problem with the door 

mechanism. At the time of the failure, several different pressure monitors inside the chamber showed the 

pressure to be in the 106 Torr range. At this temperature (-60°C) and pressure, the sublimation rate for 

water/ice is greater than the collection rate and it was unlikely that enough ice could exist on surfaces at -60°C 

to cause a problem. Colder surfaces at, -90°C or less, did have visible ice because the collection rate was 

higher than the sublimation rate. Some water/ice was probably present on mechanism components since 

there were increases in pressure when the heater was turned on to warm the housing to +25°C. However, the 

thickness of water/ice would have been angstroms and not the millimeters which would have been required to 

block the mechanism. A pressure increase was also observed with the on-board Chandra vacuum monitor 

when the heater was turned on. However, even small amounts of water/ice could result in increased pressure 

at vacuum levels of i05 to 106 Torr. The local pressure around the mechanism would have had to been an 

unlikely I 0 Torr in order for enough ice to collect to cause a problem. On a global basis then, the presence of 

ice was analytically shown to be non-credible. (However, as we have noted, locally, in a trapped interstitial 

space, the conditions can differ sufficiently to allow an iced-up seal to remain as a viable potential failure 

mode.) 

To further investigate the water/ice failure mode, an engineering unit door mechanism was purposefully iced 

up by cooling it with liquid nitrogen in ambient pressure. The mechanism and STARSYS actuator were 

covered with ice from relative humidity condensation but the running torque increase was barely measurable. 

Other tests and studies were performed, but all of our test results failed to show icing as a failure mechanism. 


I 

85 90 95 100 105 

Time (Seconds) 

110 115 120 


30 

27 

24 

21 

C 

0 

18 0C0 

I— 

15 . U 

12 

E 

9 

6 

3 

0

Further Commentary on 0-ring Adhesion Mechanisms 

During the course of this investigation various 0-ring adhesion mechanisms have been proposed as possible 

causes for the failure. Although none have been shown conclusively to occur under the failure conditions of 

the flight hardware, several remained viable possibilities. This is especially true if two or more adhesion 

mechanisms acted together synergistically. All possibilities focused on the properties of Viton adhering to the 

gold surface of the door seal under various conditions. Although this failure mode is believed to be the most 

probable failure mode, no test or analysis could conclusively prove that enough adhesion could have existed 

at TRW to prevent the door from opening. Lockheed-Martin conducted many tests with engineering hardware 

that duplicated the environmental conditions, but not the 7 months that existed before and during the failure. 

(Exponential curve-fitting of stiction values Vs time asymptote to 12.1 in-Ibs, and logarithmic fits predicted 

I 8.3 in-Ibs, tantalizingly close to the I 9 in-lb of torque measured when the flight door was first opened after the 

failure. Neither of these predicted values would create any difficulty for the ACIS actuator and mechanisms.) 

The team was unable to reproduce the failure under any simulated circumstance. 

Plans for Flight Opening 

An immense effort was expended to identify the failure mechanism prior to launch, but despite the best efforts 

of many able people, we were not able to conclusively identify the failure mechanism. In fact, it sometimes 

seemed as if the failure could not have occurred, but it had. As time went on and we approached the launch 

date without a clearly identified root cause, the team's (and program management's) approach was to 

eliminate or "mitigate" all possible remaining failure modes. The decision to open the door warm, which had 

always been the plan, would eliminate any potential issues with icing, however unlikely. 

Since 0-ring stiction seemed to be the most likely failure mechanism, even though it could not be proven, 

plans were developed to deal with it in the flight situation. First of all, it appeared that the stiction level would 

build over time, starting out at a low value when the door was initially closed, and reaching an asymptote at 

some later time. To minimize the potential stiction build-up, it was decided to open the flight door as late as 

possible in the integration process. 

Another stiction mitigation technique was developed by the Team and confirmed by Lockheed-Martin in their 

testing using engineering models of the door mechanism. It was determined that thermal cycling of the 

detector housing would lower stiction levels. Mitigation testing on the CAMSIM test unit showed that a rapid 

temperature cycle to 25°C would virtually eliminate any 0-ring adhesion which was present prior to the cycle. 

Two tests were preformed after a room temperature dwell with I atmosphere of pressure on the door for 

21 and 172 hours. Using a prediction based on previous 0-ring adhesion testing, one would estimate seal 

adhesion of 0.6 and 1.2 Nm (5.8 and 10.5 in-Ibs), respectively. After the mitigation thermal cycle (described 

above) was performed, only 0.1 Nm (0.7 in-Ibs) of adhesion were measured in both cases. 

In preparation for flight, the Chandra Flight Operations Team (FOT) developed a procedure which was 

essentially identical to the multiple-heating-cycle ground procedure, and modified the telemetry 

decommutation software and display pages to utilize the new information available on the door mechanism. 

A full end-to-end test of the new door procedure was executed on September 27, 1998 from the Chandra 

Operations Control Center in Cambridge, MA, controlling the Observatory at TRW. The ISIM was installed on 

the Observatory, which was in ambient conditions. The flight procedure utilized a library of command 

sequences that contained the actuator heater-on command, a fixed time delay, and the actuator heater-off 

command. These sequences were built with time delays of 50s to 200s in 5s intervals, and additionally in 

250s and 300s intervals. The operator could then select the command sequence with an appropriate time 

interval based on the starting temperature of the actuator, the spacecraft bus voltage, and the results of the 

previous heating cycle. On September 27, 1 998 the door opened nominally in five heating cycles. The heating 

cycles used were of 70, 85, 1 00, 1 1 0, and 250 sec. and reached maximum actuator temperatures of 

61.2, 68.9, 74.1, 76.6, and 99.9°C. The spacecraft bus voltage was 31.5 V and the input voltage at the PSMC 

was 30.9 V. 

The final pre-flight door opening took place on May 10, 1999, before Chandra was integrated with the orbiter 

Columbia. The opening went smoothly, after equalizing the pressure with the VGSE, with all data perfectly 

nominal. The instrument was commanded from the ACIS EGSE at KSC. The Observatory was in ambient 



204 

Plans for Flight Opening, Concluded 

conditions in the Vertical Assembly Building. The door was opened with 4 heating cycles of 90, 1 00, 1 1 0, and 

I 80 sec. duration's, and the actuator reached maximum temperatures of 63, 68, 72, and I 00 C. The 

spacecraft bus voltage was 30 V. 

On-Orbit Door Opening 

The Chandra X-ray Observatory was launched from KSC on July 23, 1999 in the orbiter Columbia. On August 

8, 1999 the ACIS door was opened as planned, 16 days after launch. The door opening sequence is shown in 

Figure 6. The same procedure was used to open the door in flight as had been used on the ground test at 

TRW. The only difference was the timing of the individual heating pulses. The door was opened with 

5 heating cycles of 90, 110, 125, 140 and 200 sec. and the actuator reached a maximum temperature of 

56.1, 68.9, 79.2, 87.0, and I 10.3°C. The spacecraft bus voltage was 31.0 V and the input voltage at the PSMC 

was 30.4 V. Actuator temperature, door angles and limit switch transitions were all as expected and are 

shown in the figure. With the successful door opening, the Chandra X-ray Observatory planned calibration 

observations were completed and scientific studies are now being performed as intended. The stiction 

mitigation technique we developed was essential to allow us to proceed through launch, but the actual stiction 

values seen by the Flight Unit were sufficiently low (with the KSC re-setting of our stiction-"clock") that its use 

was not required. 

ci) 

a) 

.1 

I 

120 

100 

80 

60 

40 

20 

0 


21 22 23 

Time(hrs day 220) 

Figure 6 - On-Orbit ACIS Door Opening 


Lessons Learned 

If there is one lesson to be learned, it is to always "test as you plan to fly" and not push the limits of a 

mechanism to facilitate schedule, logistics or to reduce other potential impacts to the test program. On-orbit, it 

was always planned to open the door with the camera body and 0-ring at 25 °C. During testing, if we had 

operated the door mechanism as planned, the door may have opened and we would not have had to deal with 

the long and involved failure investigation. If it had failed during test, the fault tree would have been much 

simpler since we would not have to deal with issues related to cold operation. One positive contribution, 

however, was the development of a much safer, more reliable operational procedure that ultimately would 

contribute to a successful door opening on-orbit. 

Another lesson is that either the mechanism design or operational procedure should allow for a prolonged stall 

condition to occur, due to possible binding or stiction, without mechanism component failure. In the ACIS door 

application, due to the late occurrence of the failure, the program successfully adopted an operational 

technique to preclude a failure due to a stalled condition, and also developed a technique to remove the cause 

of the stall. 

The results of the ACIS door failure investigation ultimately were inconclusive and thus very frustrating. The 

specific cause of the test failure was never positively identified, although several potential causes were left 

open as possible. As it became clear that the specific cause might not be identified, much effort was put into 

"mitigation procedures". These procedures were developed to 'mitigate" or to reduce the effect of, possible 

failure mechanisms such as 0-ring stiction. These procedures were extensively tested on ACIS door 

engineering models, and later on the flight unit itself. 

References 

I . Chandra X-Ray Observatory ACIS Failure Investigation, MSFC & Tiger Team Briefings and Final Report, 

18 June 1998 to 12 February 1999, Respectively. 

2. ACIS Door Failure Investigation and Mitigation Procedures, N. Tice, W. Podgorski and 

P. Plucinsky, 34th Aerospace Mechanisms Symposium, 1999. 

3. Relation of the Equilibrium Contact Angle to Liquid and Solid Constitution, W. Zisman, US NRL, 

Advances in Chemistry Series, No. 43, American Chemical Society, Washington, DC, 1964, 

Pgs. 1-56. 

4. Influences ofConstitution on Adhesion, W. Zisman, lnd. Eng. Chem., Vol. 55, No. 10, 1963, 

Pgs. 18-38. 

5. Liquid-Like (Transition) Layer on Ice, H. Jellinek, Dept. of Chemistry, Clarkson College of Technology, 

J. Colloid. & Interface Sci., Vol. 25, 1967, Pgs. 192-205. 

6. Studies of Ice Adhesion, US. Naval Applied Science Laboratory, M. Landy & A. Freiberger, 

J. Colloid. & Interface Sci., Vol. 25, 1967, Pgs. 231-244. 

7. Ice Adhesion to Hydrophilic and Hydrophobic Surfaces, W. Bascom, R. Cottington, & C. Singleterry, 

Chemistry Division, US NRL, J. Adhesion, Vol. 1, Oct. 1969, Pgs. 246-263. 

8. The Friction and Creep of Polycrystalline Ice, P. Barnes, D. labor, & J. Walker, Surface Physics, 

Cavendish Laboratory, Cambridge, Proc. Roy. Soc. London, Vol. A342, 

l2Augustl97l, Pgs. 127-155. 

9. Elastic Properties ofAmorphous and Crystalline Ice Films, J. Hessinger, B. White, Jr., & 

R. PohI, Laboratory of Atomic and Solid State Physics, Cornell University, Planetary and 

Space Science, Vol. 44. No. 9. Sept. 1996, Pgs. 937-944. 

I 0. Physical Chemistry of Surfaces, Third Edition, Arthur Adamson, Dept. of Chemistry, Univ. of Southern 

Calif., J. Wiley & Sons, Pgs. 454 & 458. 

I I . On Stickiness, C. Gay & L. Leibler, Physics Today, November 1 999, Pgs. 48-52. 

12. Astrophysics CCD Imaging Spectrometer (ACIS) Failure Investigation Report, Adhesion Mitigation 

Methodology Final Report, S. Anderson, Lockheed Martin Astronautics Flight Systems, 

28 January 1999.

ACIS door failure investigation: technical details and mitigation ...

Transform your PDFs into Flipbooks and boost your revenue!

Delete template?

Save as template ?